<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Tool-Use on skulldc</title><link>https://skulldc.net/tags/tool-use/</link><description>Recent content in Tool-Use on skulldc</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Fri, 06 Nov 2026 10:26:06 +0000</lastBuildDate><atom:link href="https://skulldc.net/tags/tool-use/index.xml" rel="self" type="application/rss+xml"/><item><title>DPDM routes agent decisions through one frozen conditional predictor instead of prompting</title><link>https://skulldc.net/2026/11/dpdm-routes-agent-decisions-through-one-frozen-conditional-predictor-i/</link><pubDate>Fri, 06 Nov 2026 10:26:06 +0000</pubDate><guid>https://skulldc.net/2026/11/dpdm-routes-agent-decisions-through-one-frozen-conditional-predictor-i/</guid><description>A paper proposes DPDM, which pairs a language model encoder with a frozen pretrained predictor so agents learn from deployment history without any weight updates.</description></item><item><title>ToolSpace builds agent training environments from real MCP tool traces</title><link>https://skulldc.net/2026/11/toolspace-builds-agent-training-environments-from-real-mcp-tool-traces/</link><pubDate>Wed, 04 Nov 2026 10:27:47 +0000</pubDate><guid>https://skulldc.net/2026/11/toolspace-builds-agent-training-environments-from-real-mcp-tool-traces/</guid><description>ToolSpace mines tool dependencies from real MCP execution traces, turns them into environments and checked tasks, and reports large gains for open models on long-horizon tool use.</description></item></channel></rss>